Viz for lidar pre processing base implementation

[jasonkdeng]

📑 Description

Stitched viz.py into existing Open3D pipeline to reduce the brittleness of viz for lidar preprocessing. Currently may still require some fixes on the lidar preprocessing docker image, but no new bugs. Attached below is a video of the html interface for viz.

📹 (Optional) Video Demo of Changes

Recording.2026-07-09.183743.mp4

✅ Checklist

• My code builds and runs locally without warnings
• I added/updated tests if needed
• I updated documentation / comments
• I listed any breaking changes in the “Notes” section

📝 Notes for reviewers

./watod run ingest --bag /data/bags/<the bag>

./watod run lidar_preprocessing --bag NuScenes_v1_0_mini_scene_0655 --chunk 0000 --force --no-auto-reduce

./watod run lidar_preprocessing viz \
  --bag NuScenes_v1_0_mini_scene_0655 \
  --chunk 0000 \
  --backend html \
  --open

[brianzheng206]

good stuff, couple of small things

[brianzheng206]

some of this is kinda slop, I also think if the html backend is able to provide us everything we need we can ditch open3d, and have that as default.

[jasonkdeng]

that's really fair. I moved open3d to just be a fallback and added the point classification to html directly

[brianzheng206]

any reason why we are changing this?

[jasonkdeng]

Lowkey added to make sure that the project pytest dependency is always aligned, but you're right. The previous written in hardcode is stable and will be good. I'll fix that

[brianzheng206]

ideally dont change this file, would you be able to define everything with cli.py? If we settle on one viz I think we wont even need this

[jasonkdeng]

ye

[brianzheng206]

not sure if we documented anywhere either but run pre-commit run --all-files to pass the precommit check

[jasonkdeng]

not sure if we documented anywhere either but run pre-commit run --all-files to pass the precommit check

it's documented in the readme